Geoscience Reference
In-Depth Information
3. Parameter Estimation and Flood Simulations
The nine floods with more than 400 m
3
/s peak discharge were selected from
the discharge data since 1978. For each flood, spatial distributions of hourly
rainfall with 3 km gird resolution were generated from the ground gauged
rainfall measurements using the nearest-neighbor method; then the five
model parameters above mentioned that define the stage-discharge rela-
tionship are determined. The initial water depth at each slope segment is
determined from the initial river discharge at the outlet assuming a steady-
state condition. Table 1 summarizes the results of parameter identifica-
tions. To evaluate appropriateness of the simulated discharges, the peak
discharge ratio and the Nash-Sutcliffe eciency are used. The evaluation
results suggest that the identified model parameter sets are classified into
three groups: group 1 with the 1993, 1981, and 1982 flood; group 2 with
the 1985, 1983, and 1979 flood; and group 3 with the 1989, 1990, and 2004
flood. The parameter sets in group 1 have tendency to overestimate and
group 3 to underestimate floods; group 2 shows both tendencies. The esti-
mated peak discharges are widely distributed from 2,500 to 4,200 m
3
/s, and
it is larger than the estimated peak discharge 2,400 m
3
/s obtained by the
Construction Ministry from hydraulic river flow simulations with the high
flood stage marks.
4. Discussion
Within a group, the values of model parameters are close, and floods sim-
ulated with any parameter sets show good scores of the peak ratio and the
Nash eciency. The clear difference among the groups is that the value of
β
is larger and the capacity of non-capillary layer
d
a
-
d
m
is smaller in group 1
as compared to group 3. The difference of parameter values represents the
difference of hydrologic characteristics. In group 1, rainfall stored in soil
layer flows quite slow, therefore at the beginning of floods, river discharge
is insensitive to rainfall intensity. Then the soil layer is easily saturated and
once the water depth at slope segments exceeds the capacity of the capillary
soil layer, river discharge rises up suddenly. On the contrary, the parameter
sets in group 3 tends to show opposite characteristics that a hydrograph
rises up from the beginning of rainfall and its peak discharge is smaller
when the same rainfall is given to the runoff model with parameter sets in
group 1. Group 2 shows the middle feature of groups 1 and 3.
The question is why the difference is observed in the same catchment.
Each group includes various scales of floods and there are no distinguished
